1996fanrui opened a new pull request, #28663: URL: https://github.com/apache/flink/pull/28663
This PR depends on https://github.com/apache/flink/pull/28651, https://github.com/apache/flink/pull/28652, https://github.com/apache/flink/pull/28659, https://github.com/apache/flink/pull/28660, https://github.com/apache/flink/pull/28661 and https://github.com/apache/flink/pull/28662 Please only review the last 6 commits (prefixed [FLINK-40081]). ## What is the purpose of the change [FLINK-40081] Enable the spilling backend for checkpointing during recovery. It's part 7 (the final part) of FLINK-38544 (checkpointing during recovery of unaligned checkpoint state). This PR swaps the memory backend for the disk backend behind `execution.checkpointing.during-recovery.enabled`: recovered channel state is fetched (and rescale-filtered) into spill files, the drainer pushes it into the recovery-state channels while serving as the recovery-checkpoint trigger, and checkpoints taken during recovery snapshot the not-yet-drained slice. The transitional memory pieces (in-memory trigger, conversion-time push, v1 filtering handler, and the unbounded heap fallback with its OOM risk) are deleted. ## Brief change log - [FLINK-40081] Switch flag-on recovery to fetch->drain; retire the in-memory backend - [FLINK-40081][network] Remove the unbounded heap fallback from RecoveredInputChannel#requestBufferBlocking - [FLINK-40081] Harden RecordFilterContext for minimal and batch environments - [FLINK-40081][test] Add UnalignedCheckpointDuringRecoveryITCase - [FLINK-40081][test] Add RescaleFilterLargeRecordOOMRegressionITCase - [FLINK-40081][test] Add CdrRecoveryRaceITCase for the non-source recovery race ## Verifying this change - New IT cases: UnalignedCheckpointDuringRecoveryITCase, RescaleFilterLargeRecordOOMRegressionITCase (the OOM scenario the disk backend eliminates), CdrRecoveryRaceITCase - New unit tests: ChannelIOExecutorDrainSubmissionTest, RecoveredInputChannelRequestBufferBlockingHeapFallbackRemovedTest, RecordFilterContextTest - The randomized ITCase fleet runs `execution.checkpointing.during-recovery.enabled` in ~50% of runs and stays green across the backend swap ## Does this pull request potentially affect one of the following parts: - Dependencies (does it add or upgrade a dependency): no - The public API, i.e., is any changed class annotated with `@Public(Evolving)`: no - The serializers: no - The runtime per-record code paths (performance sensitive): no - Anything that affects deployment or recovery: JobManager (and its components), Checkpointing, Kubernetes/Yarn, ZooKeeper: no - The S3 file system connector: no ## Documentation - Does this pull request introduce a new feature? no -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: [email protected] For queries about this service, please contact Infrastructure at: [email protected]
